Production and Characterization of Novel Recombinant Adeno-Associated Virus Replicative-Form Genomes: A Eukaryotic Source of DNA for Gene Transfer

Conventional non-viral gene transfer uses bacterial plasmid DNA containing antibiotic resistance genes, cis-acting bacterial sequence elements, and prokaryotic methylation patterns that may adversely affect transgene expression and vector stability in vivo. Here, we describe novel replicative forms of a eukaryotic vector DNA that consist solely of an expression cassette flanked by adeno-associated virus (AAV) inverted terminal repeats. Extensive structural analyses revealed that this AAV-derived vector DNA consists of linear, duplex molecules with covalently closed ends (termed closed-ended, linear duplex, or “CELiD”, DNA). CELiD vectors, produced in Sf9 insect cells, require AAV rep gene expression for amplification. Amounts of CELiD DNA produced from insect cell lines stably transfected with an ITR-flanked transgene exceeded 60 mg per 5×109 Sf9 cells, and 1–15 mg from a comparable number of parental Sf9 cells in which the transgene was introduced via recombinant baculovirus infection. In mice, systemically delivered CELiD DNA resulted in long-term, stable transgene expression in the liver. CELiD vectors represent a novel eukaryotic alternative to bacterial plasmid DNA.


Introduction
Non-viral gene transfer typically uses bacterial plasmids to introduce foreign DNA into recipient cells. In addition to the transgene of interest, such DNAs routinely contain extraneous sequence elements needed for selection and amplification of the plasmid DNA (pDNA) in bacteria, such as antibiotic resistance genes and a prokaryotic origin of replication. Also, bacterial DNA preparations often contain endotoxin that can reduce gene transfer efficiency [1]. Ideally, DNA for non-viral gene transfer would contain only the gene of interest, be devoid of prokaryotic modifications that can trigger an innate immune response [1,2,3,4], be in an exonuclease-resistant form, and lack detectable endotoxin contamination. Transgene persistence is also highly desirable for many gene therapy applications. We have observed that transgenes flanked by inverted terminal repeats (ITRs) from adeno-associated virus type 2 (AAV) can be amplified to high copy number in Spodoptera frugiperda (Sf9) cells in the presence of a recombinant baculovirus expressing the AAV replication proteins, Rep78 and Rep52. DNA produced in this manner consists solely of a transgene of interest with flanking AAV ITRs, and, since it is produced in eukaryotic cells, is devoid of prokaryotic DNA modifications and bacterial endotoxin contamination. Structural characterization revealed the ITR-flanked transgenes as exonuclease-resistant, linear DNA molecules with covalently closed ends. We have termed these closed-ended, linear duplex molecules ''CELiD'' DNA.
CELiD DNA is produced in Sf9 cells upon co-infection with two separate baculovirus expression vectors (BEVs): the first BEV bearing an ITR-flanked transgene, and a second BEV (designated Bac-Rep) encoding AAV non-structural proteins (Rep78 and Rep52) essential for ITR-mediated DNA replication. Alternatively, CELiD DNA can be rescued and amplified upon Bac-Rep infection of clonal Sf9 cell lines bearing a stably integrated rAAV vector genome. The structure of replicated ITR-flanked transgene DNA was analyzed using native and denaturing agarose gel electrophoresis, restriction endonuclease size mapping, atomic force microscopy (AFM), and exonuclease sensitivity assays. The results of these orthogonal assays established that the predominant structure of the replicated vector DNA is closed-ended, linear duplex monomers and dimers.

Ethics statement
The University of North Carolina at Chapel Hill Institutional Animal Care and Use Committee (IACUC) approved the animal study protocols.
Establishing stable Sf9/ITR-GFP cell lines for high-yield CELiD-GFP DNA production Sf9 cells were transfected with pFBGR-bsd using Cellfectin Transfection Reagent (Life Technologies, Corp.). At three days post-transfection, antibiotic-resistant cells were selected by the addition of blasticidin-S HCl (50 mg/ml) (Life Technologies, Corp.) to the growth medium. After two weeks in selective medium, blasticidin-resistant (bsd r ) clones were derived by singlecell dilution or direct colony transfer techniques. The bsd r clones were expanded in insect cell culture medium supplemented with 10% fetal bovine serum (FBS) (HyClone) and blasticidin-S HCl (10 mg/ml) for 2 to 3 additional passages, then returned to serumfree medium with 10 mg/ml blasticidin-S HCl. After an additional 12 passages, blasticidin-S HCl was omitted from the medium and the cell lines were expanded for analysis.
For functional screening, clonal Sf9/ITR-GFP cell lines were infected (MOI = 5) with a recombinant baculovirus, Bac-Rep, expressing the AAV type 2 Rep78 and Rep52 proteins [5] and analyzed for induced GFP expression. Clonal Sf9/ITR-GFP cells with the highest levels of GFP fluorescence were expanded for CELiD-GFP DNA preparation. Production and purification of CELiD-GFP DNA from a clonal Sf9/ITR-GFP cell line Clonal Sf9/ITR-GFP cells were seeded at 2610 6 cells/mL and infected with Bac-Rep (MOI = 1 to 3). Cell viability and diameter were monitored daily until the cell diameter increased to 18-20 mm (uninfected cell diameter 14-15 mm), indicating that the cells were in the late stages of the viral infection. Extrachromosomal DNA was extracted from the Bac-Rep-infected, Sf9/ITR- The clonal Sf9/ITR-GFP cell line contains a stably integrated copy of pFBGR-bsd. Uninfected cells or cells infected with wild-type baculovirus (Bac-AcNPV) lack detectable GFP expression as determined by FACS analysis. (upper row, 0% GFP positive cells). Addition of Bac-Rep inoculum (0.5%, 1.0%, and 1.5%; v:v) resulted in a dose-dependent increase in the number of GFP-positive cells (lower row). Images were obtained 3 days after infection. Magnification, 10x objective used for all images. (B) Quantitative analysis of GFP induction as a function of Bac-Rep dose. Cells exposed to increasing doses of Bac-Rep were harvested on day 3 post-infection. Fluorescent emission intensities were assessed from equivalent amounts of cell lysates (80 mg protein), using the fluorescent reader function of a real-time thermocycler (excitation 450-490 nm; emission 515-530 nm) (upper panel), *** indicates t-test significance (p,0.001). The relative fold-increase in GFP fluorescence is indicated by the values next to each bar. Protein concentrations were determined in the lysates and approximately 120 mg of each sample was fractionated electrophoretically using SDS-polyacrylamide gels. Proteins were electroblotted onto nitrocellulose membranes for western blot detection of GFP, Rep52, gp64, and GAPDH (used as a protein loading and transfer standard, lower panel). Relative levels of protein are indicated by the values above the protein band. (C) Analysis of the increase of GFP-vector DNA in response to Bac-Rep infection. Low molecular weight DNA was recovered from the cells and the quantity of GFP vector genomes determined using real-time qPCR (upper panel). Both the uninfected cell control and the wild-type baculovirusinfected cell control lysates produced a relatively low PCR signal (150 and 175 copies per cell, respectively). The GFP vector DNA copy number increased to 21087 copies per cell with 0.5% (v:v) Bac-Rep. At 1% (v:v) Bac-Rep, the copy number increased to 28862 copies per cell and 1.5% (v:v) Bac-Rep increased the copy number to 38286 per cell. The statistical significance was determined by Student's t-test. ***, p,0.001. Lower panel -The samples were analyzed by agarose-ethidium bromide gel electrophoresis (1% agarose). No detectable CELiD DNA was recovered from uninfected cells. A band .10 kb appears in lysates from all baculovirus-infected cells. (D) Time-course of GFP fluorescence. Sf9/ITR-GFP cells were inoculated with either wtBac (Bac-ACNPV) or Bac-Rep (0.5%, 1%, and 1.5% (v:v)). GFP fluorescence was measured for the indicated cellular lysate at 1, 2, and 3 days post-infection. doi:10.1371/journal.pone.0069879.g002 GFP cells using a commercially available plasmid isolation kit (Qiagen, Valencia, CA, USA). CELiD production was monitored by agarose gel electrophoresis and ethidium bromide staining of extrachromosomal DNA. The presence of a 2.7 kb band after Bac-Rep infection indicated that the ''proviral'' rAAV-GFP genome was rescued from chromosomal DNA and amplified as an episomal element.
CELiD DNA production in parental Sf9 cells CELiD DNA was produced in parental Sf9 cells by coinfection with two separate baculovirus expression vectors (BEV): Bac-Rep and a second BEV bearing an ITR-flanked transgene, such as Bac-GFP or Bac-LacZnls. Infected Sf9 cells were harvested once the mean cell diameter increased by 4-5 mm and the percent viability decreased to 80-90%. CELiD DNA was isolated using a commercially available plasmid purification kit (Qiagen).

Construction of pFB-TBG-GFP and baculovirus production
A thyroxine-binding globulin promoter (TBG)-GFP cassette (1797 bp) was isolated from pENN AAV TBG PI-eGFP (kindly provided by Dr. Charles P. Venditti, NHGRI, NIH) following digestion with NheI and PspXI. The ca. 1800 bp TBG-eGFP fragment was transferred into the backbone of pFBGR after the GFP transcription cassette was removed from pFBGR by digestion with SpeI and XhoI (retaining the AAV2 ITRs and transposase Tn7 recombination signals). The resulting pFB-TBG-GFP (7074 bp) plasmid consists of the AAV2 ITRs flanking the TBG promoter and GFP open reading frame. Recombinant baculovirus stocks were generated using the Bac-to-Bac System (Life Technologies, Corp.) with modifications to the plasmids previously described [5]. Briefly, competent E. coli DH10Bac cells were transformed with pFB-TBG-GFP. Following selection on triple antibiotic agar plates (containing 50 mg/ml kanamycin, 7 mg/ml gentamicin, 10 mg/ml tetracycline, and X-gal 100 mg/ ml), several white colonies were picked and expanded in small volume cultures. Bacmid DNA was extracted and used to transfect Sf9 cells. Three days after transfection, baculoviruscontaining supernatant (passage 1 or ''P1'') was expanded two more rounds by adding infected-cell supernatant to fresh Sf9 cells (1:100; v:v). Infected cells were cryopreserved at P3, and served as a source of infectious baculovirus as previously described [6].

Denaturing agarose gel electrophoresis
Alkaline (denaturing) agarose gels were prepared by dissolving agarose (SeaKem LE agarose, BioWhittaker Molecular Applications, Rockland, ME, USA) in boiling water and cooling to near gelling temperature. Alkaline gel solution concentrate (10x) was added to a 1x final concentration just before the gels were poured. The 1x alkaline gel running buffer and gel solution composition is: 50 mM NaOH, 1 mM EDTA. Using 10x concentrate, DNA samples were adjusted to 1x alkaline gel solution conditions and supplemented with 2.5% Ficoll-400 and 0.05% bromophenol blue. Electrophoresis was performed at constant voltage (30-60 V) for 3-6 hr.

Time-course analysis and requirements for CELiD rescue and amplification in Sf9 cells
Clonal Sf9/ITR-GFP cells were inoculated with various amounts of Bac-Rep stock. Periodically, cells were harvested and extrachromosomal DNA was recovered using a commercially available DNA isolation kit (Qiagen Plasmid Mini Kit). Extracted DNA was examined by either agarose gel electrophoresis or by PCR with GFP-specific primer pairs for quantitative determination of CELiD DNA amounts.
Prior to applying the DNA samples, AFM mica substrates were modified with aminopropylsilatrane (APS). APS solution (,150 nM in 50 ml) was applied to freshly cleaved mica disks (12 mm diameter) for 30 minutes at ,20uC. The disks were then rinsed thoroughly with dH 2 O and dried in a nitrogen stream. DNA solutions of appropriate concentrations (typically, on the order of 1 mM base-pair concentration) were applied to the APStreated mica and incubated approximately 10 min at room temperature. The samples were then gently rinsed with dH 2 O and dried in an argon stream before imaging.
The effects of different solvents on the topological features of the sample DNA were examined by AFM using: 1. Buffered saline (500 mM NaCl and 20 mM HEPES, pH 7.4), 2. 50% formamide (sample heated at 60uC for 30 minutes to denature the DNA), and 3. Water (dH 2 O).
Samples were imaged with an AFM instrument (MultiMode-PicoForce AFM with Nanoscope V controller, Bruker Nano, Santa Barbara, CA, USA). Imaging was performed in oscillating mode (tapping) under ambient conditions using silicon cantilevers with nominal stiffness of ,42 N/m and resonance frequencies in the 300 kHz range (OTESPA by Olympus or TESP-SS by Bruker Nano). Imaging data were preprocessed using Nanoscope v7.31. DNA lengths were measured using algorithms developed in conjunction with image processing software based on NIH ImageJ (http://rsb.info.nih.gov/ij/), Matlab (Mathworks, Natick, MA, USA) and Origin (OriginLab, Northampton, MA, USA).
The lengths of extended DNA chains with clearly discernible paths were determined by contour tracing. For overlapping and entangled chains, we developed new algorithms that made use of the volume information contained in the AFM topographic images. Volumes were measured by adding pixel heights using the Particle Analysis module of NIH ImageJ.

CELiD DNA expression in vitro
Transfections were conducted using either GFP or nuclearlocalized b-galactosidase (LacZnls) reporter cassettes. HEK 293 cells were seeded in 6-well tissue culture plates one day before transfection to achieve 90% confluence at the time of use. Transfections were performed using a commercially available liposome-based transfection reagent (LipofectamineH 2000, Life Technologies, Corp., Carlsbad, CA, USA) following the manufacturer's suggested protocol. For GFP expression assays, HEK 293 cells were transfected using 4 ml of lipid reagent per well and equivalent copy numbers of reporter DNA molecules: CELiD-GFP DNA (1.48 mg per well) or plasmid-GFP DNA, pFBGR (4 mg per well). At day 6 post-transfection, cells were examined for GFP expression by fluorescence microscopy.
For LacZ expression assays, HEK 293 cells were seeded in 6well tissue culture plates and transfected with equal weight amounts of CELiD-LacZnls (2 mg) or pFB-CMV-LacZnls (2 mg) using 3 ml of lipid reagent per well. Three days post-transfection,  Transfected livers were harvested and processed at days 1, 3, and 7, as well as 5 weeks post-injection. Histological samples were sectioned and b-galactosidase (b-gal) activity was detected by staining with the chromogenic substrate, X-gal. Briefly, cells were rinsed with PBS, and then fixed in 2% formaldehyde, 0.1% glutaraldehyde in PBS for 2 minutes. After washing the cells three times with PBS, freshly made X-gal staining buffer (1 mg/ml X-gal, 2 mM MgCl2, 5 mM potassium ferrocyanide, and 5 mM potassium ferricyanide in PBS, warmed to 50uC) was added, and the cells were incubated at 37uC for 30 min to 24 hr in the dark. After incubation, samples were washed three times (each for 5 min) in room temperature PBS, and examined by light microscopy.

Hydrodynamic injection of DNA and in vivo gene expression detection
Outbred male ICR mice at 2 months of age were used for hydrodynamic DNA injection via the tail vein according to the method described by Liu et al. [7]. Each mouse was injected with 1.7 ml of saline containing circular plasmid DNA or CELiD DNA. For b-galactosidase activity, 10 mg of pCMV-LacZnls or 1 mg of CELiD-LacZnls DNA was used. Mice were sacrificed periodically (1, 3, and 7 days, as well as 5 weeks, post-injection) and liver tissue collected and snap-frozen in liquid nitrogen for X-gal staining of tissues. The liver tissues were cryo-sectioned (15 mm thickness), fixed and X-gal stained according to a standard method [8]. Samples were counter-stained with eosin, dehydrated with ethanol, and mounted for imaging.
For TBG-GFP gene expression in vivo, 10 mg of either CELiD-TBG-GFP or circular plasmid pTBG-GFP in 1.7 ml saline was injected via the tail vein. Mice were sacrificed at 1, 3, and 10 wks post-injection, and livers samples were processed as above. GFP expression within the liver sections was detected by fluorescence microscopy. N = 3 to 4 animals per experimental time point.

Statistical analysis
Quantitative results are expressed as the mean plus or minus (6) the standard deviation (SD). Statistical significance analysis of the data was performed using an unpaired two-sample equal variance and two-tail distribution Student's t-test (t-test). A p-value of less than 0.05 was considered statistically significant as indicated by asterisks: * p,0.05, ** p,0.01, *** p,0.001.

Characterization of replicative-form vector DNA produced in stably-transfected Sf9 cell lines
A plasmid (pFBGR-bsd) containing a blasticidin-S deaminase selectable marker linked to an AAV ITR-flanked green fluorescent protein (GFP) reporter gene under the regulatory control of the CMV IE promoter and the baculovirus p10 promoter is shown in Fig. 1. Sf9 cell lines were produced by transfection with pFBGR-bsd and selection for stable blasticidin-S deaminase expression. Infection of blasticidin-resistant cells with a recombinant baculovirus expressing the AAV2 rep gene (Bac-Rep) induced GFP expression and amplification of the AAV ITR-flanked GFP transgene in a dose dependent manner (Fig. 2). In contrast, infection with wild-type baculovirus (Bac-AcNPV) had no effect on GFP expression. Characterization of the relative amounts of the Rep and GFP proteins by western blot analysis combined with quantification of GFP-associated fluorescent intensity indicated a direct correlation between Rep protein expression and the induction of GFP activity in infected cells (Fig. 2B).
To test the hypothesis that Rep protein expression results in rescue and amplification of the integrated GFP vector genome, extrachromosomal DNA was extracted from stably transfected Sf9/ITR-GFP cells and analyzed using quantitative PCR (qPCR) (Fig. 2C, upper panel) and agarose gel electrophoresis (Fig. 2C,  lower panel). A unit-length 2.7 kb band is apparent in the Bac-Rep-infected cells, but not in cells infected with wild-type (wt) baculovirus (Bac-AcNPV). Using qPCR, it was determined that the copy number of the GFP target sequence increased to greater than 15,000 copies per cell, in proportion to the amount of Bac-Rep added to the culture (Fig. 2C, upper panel).
During a time-course experiment, higher doses of Bac-Rep led to a proportional increase of GFP fluorescence observed on both day 2 and day 3 post-infection. By day four, the fluorescence intensity reached a maximum value regardless of the initial multiplicity of infection (Fig. 2D). The rate of accumulation of the GFP vector DNA was analyzed using samples of Sf9/ITR-GFP cells infected with 1% (v:v) of baculovirus-containing supernatant (Fig. 3A). The GFP vector genome copy number increased non-linearly over three days. At 168 hr post-infection, the amount of GFP vector genomes decreased, presumably due to loss of cell viability. The staining intensities of the monomer and dimer replicative-form DNAs (2.7 and 5.4 kb, respectively) (Fig. 3A, lower panel) indicate that the peak level of amplification occurred at the 55 hr time point. The rate of Rep 78 and Rep 52 expression reached a maximum at 77 hr post-infection, then declined at 168 hr (17% of 77 hr Rep expression) (Fig. 3B). The amount of GFP detected by western blot decreased slightly from 77 to 168 hr (69% of 77 hr GFP expression) (Fig. 3B). stained to detect b-galactosidase activity (indicated by dark blue nuclei). Samples were counterstained with eosin. (C) CELiD-LacZnls and plasmid-LacZnls copy number per diploid cell in transfected livers. The DNA copy number was normalized based on the PCR quantification of the endogenous mouse glucagon gene, n = 3 to 4. (D) Comparative long-term transgene expression from constructs bearing a liver specific thyroxine-binding globulin (TBG) promoter. CELiD-TBG-GFP or plasmid pTBG-GFP gene expression in liver section 10 weeks post-hydrodynamic tail vein injection. (E) CELiD-TBG-GFP and plasmid pTBG-GFP DNA copy number per diploid cell in transfected livers. The same amount of DNA (10 mg) was administrated by hydrodynamic injection. Statistical analysis by TTEST. ** indicates P,0.01. doi:10.1371/journal.pone.0069879.g008 Characterization of CELiD DNA Extrachromosomal DNA was extracted from Bac-Rep-infected Sf9/ITR-GFP cells and resolved on either native or denaturing agarose gels. Under native conditions, the extrachromosomal DNA appeared predominantly as monomeric (2.7 kb) and dimeric (5.4 kb) forms of the vector genome, within a ladder of vector genome multimers (Fig. 4A). Under denaturing conditions, however, the extrachromosomal DNA presented predominantly as approximately 5.4 kb and 10.8 kb bands (Fig. 4B). The change in mobility under denaturing conditions suggested that the monomeric and dimeric vector genomes are duplex molecules with at least one covalently closed end.
To determine whether the ends of the linear ITR-flanked GFP transgene are covalently closed, extrachromosomal DNA recovered from Sf9/ITR-GFP cells was treated with the restriction endonuclease AgeI and resolved using either native or denaturing agarose gel electrophoresis (Fig. 4C) (see also Fig. S1). Following digestion with AgeI, the extrachromosomal DNA resolved as two major bands, of 1.8 kb and 0.9 kb, and two minor bands of 3.6 kb and 1.7 kb (Fig. 4C, right). The 1.8 kb and 0.9 kb bands are as expected for digestion of a vector genome monomer with AgeI, whereas the 3.6 kb and 1.7 kb bands are consistent with head-tohead and tail-to-tail concatamers of multimeric vector genomes (Fig. 4C, left). Under denaturing conditions, the AgeI digestion products appeared as just two bands (of 3.6 kb and 1.7 kb). Lack of the 1.8 kb and 0.9 kb bands suggests that the replicated vector genomes are covalently closed at both ends of the DNA duplex. Use of other restriction endonucleases produced results consistent with this model (Fig. S1).
As second indication of whether the ends of the vector DNA are covalently closed, isolated extrachromosomal DNA was subjected to digestion with either E. coli exonuclease I (ExoI) or exonuclease III (ExoIII). ExoI degrades single-stranded DNA in the 39 to 59 direction [9]. ExoIII progressively degrades duplex DNA in a 39 to 59 direction from 39-recessed or blunt-ended DNA molecules (substrates with 39 protruding ends greater 4 nt are resistant) [10,11]. Figure 5 shows that CELiD DNA and circular singlestranded wX174 DNA were resistant to ExoI digestion, while a single-stranded, synthetic oligonucleotide or endonuclease HaeIIIdigested wX174 DNA were susceptible. As shown in Fig. 5A, native CELiD DNA was resistant to ExoIII digestion; however, after treatment with restriction enzyme NaeI, the vector DNA became susceptible to ExoIII activity. Together with the denaturing agarose gel electrophoresis results, these observations are consistent with a replicative-form DNA bearing a closed-ended DNA duplex structure.

Atomic Force Microscopy of CELiD -vector DNA
Atomic force microscopy (AFM) was used to visualize purified CELiD DNA (Fig. 6). The 2.7 kb CELiD-GFP vector genome appeared as an extended, linear duplex molecule with an estimated mean contour length of 2701 (6607) base pairs (Fig. 6A). Results of alkaline agarose gel electrophoresis analysis predict that, upon denaturation, the 2.7 kb CELiD-GFP monomer should occur as an approximately 5400 base-long, singlestranded, circular molecule. Although maintaining CELiD DNA in a single-stranded state proved difficult, several predominantly single-stranded DNA molecules with closed contours were observed under denaturing conditions (Fig. 6B). The lengths of extended DNA chains with clearly discernible paths were determined by contour tracing and were consistent with a covalently closed, single-stranded monomer.
AFM analysis of the 5.4 kb CELiD-GFP DNA dimer revealed an interesting conformation. The termini of the dimer appear as compacted, tortuous structures with a similar, but less extensive, structure centrally located (Fig. 6C). The contours of these molecules could not be traced; however, the lengths were estimated using total molecular volumes. The calculated DNA lengths averaged 5480 bp (6310 bp; n = 23) in agreement with the predicted size of 5336 bp. The positions of the convoluted regions of the dimer DNA correspond to the locations of the AAV ITRs. The AAV2 Rep78 protein, required for CELiD-vector DNA production, has sequence specific DNA binding activity. Multimers of Rep78 have been shown to non-covalently interact with a motif repeated in the AAV ITR [12,13]. Western blots of nuclease-treated CELiD-vector DNA preparations detected the presence of both Rep78 and Rep52 (Fig. S2). To investigate whether Rep proteins, or other proteins, binding to the DNA contributed to the convoluted structures, the ionicity was increased to potentially disrupt non-covalent protein-DNA interactions. In the presence of 0.5 molar NaCl, the 5.4 kb CELiD-GFP DNA transitioned to an extended conformation (Fig. 6D). The lengths of several such relaxed DNA molecules were measured by contour tracing (53376118 bp; n = 18), and were consistent with the lengths estimated for the convoluted forms using the volume per unit length estimates. The measured lengths agree with the predicted size of the duplex dimer (5336 bp).

CELiD -vector DNA expression in vitro and in vivo
The ability of CELiD DNA produced in Sf9 cells to direct transgene expression in vitro was determined using an ITR-flanked green fluorescent protein reporter cassette (CELiD-GFP, Fig. 7A) or an ITR-flanked LacZnls (nuclear-localized b-galactosidase) reporter cassette (CELiD-LacZnls, Fig. 7B). Equivalent copy numbers of parental GFP plasmid and CELiD-GFP DNA molecules were used in the in vitro experiments. FACS analysis of cells transfected with CELiD-GFP or pFBGR (a GFPexpressing plasmid) showed that the population of positive cells and expression levels were similar: 39.6% and 31.1% GFP positive cells and 3656 and 2442 relative fluorescent units for CELiD and pFBGR-transfected cells, respectively (Fig. 7A). Using equal weight amounts of DNA, LacZ expression was higher for plasmid DNA than CELiD DNA in transiently transfected HEK 293 cells (Fig. 7B).
In vivo gene expression was assessed in outbred ICR male mice using purified CELiD-LacZnls DNA administered by hydrodynamic tail vein injection (Fig. 8). Livers from mice injected with CELiD-LacZnls DNA (1 mg) had detectable levels of b-galactosidase activity at 1 day after injection, and the level of expression remained relatively constant through day 7. LacZnls activity could still be detected at 5 weeks post-injection in CELiD-LacZnls liver section (Fig. 8A). Plasmid pCMV-LacZnls DNA (10 mg) demonstrated robust expression at day 1, but quickly diminished over the course of the experiment. Although more plasmid DNA was injected than CELiD-LacZnls DNA, bgalactosidase activity in the pCMV-LacZnls-treated liver was not detectable at 5 weeks (Fig. 8B); however, the CELiD-LacZnlstreated liver still had detectable LacZnls expression at a similar time point. Quantitative PCR analysis of vector DNA copy number in transfected liver tissue showed that more CELiD-LacZnls DNA was retained in the liver at 5 weeks post-injection compared to plasmid DNA, and also showed a slower rate of decline in copy number over time (Fig. 8C). We postulate that the decline of LacZ expression over time is due primarily to shutoff of the CMV promoter in the liver. To address this hypothesis, a GFP reporter gene under the control of the thyroxine-binding globulin (TBG) promoter, which demonstrates prolonged activity in hepatocytes [14,15,16], was used to assess long-term gene expression in the mouse liver. CELiD-TBG-GFP and the equivalent plasmid pTBG-GFP DNA (10 mg each) were administrated to outbred ICR mice by hydrodynamic tail vein injection. At 10 weeks post-injection, no GFP expression was evident in the plasmid-treated mice, while GFP expression in livers of CELiD-TBG-GFP-treated mice was apparent (Fig. 8D). The CELiD-TBG-GFP expression pattern was predominantly centrilobular, resembling expression patterns observed in recombinant AAV-treated mouse livers [17]. Quantitative analysis of DNA copy numbers of CELiD-TBG-GFP and plasmid pTBG-GFP DNA showed very similar levels at 1 wk post-injection (Fig. 8E). CELID-TBG-GFP DNA copy number remained at 91% at 3 wks and 65% at 10 wks post-injection relative to the amount of vector DNA present at the 1wk time point, while plasmid pTBG-GFP DNA copy numbers decreased to 39% at 3 wks and only 1.8% at 10 wks post-injection relative to the 1 wk pDNA amount. Therefore, the amount of vector DNA in the liver of CELiD-TBG-GFP-recipient mice was 36-fold higher than those receiving equivalent plasmid DNA. The differences between CELiD-TBG-GFP and plasmid pTBG-GFP DNA copy numbers in transfected livers at 3 wks and 10 wks post-injection demonstrated statistical significance (P,0.01). The CELiD-TBG-GFP in vivo data revealed that the levels of CELiD DNA in mouse liver remained nearly unchanged during a 10 wk period.

Discussion
Non-viral gene delivery circumvents certain disadvantages associated with viral transduction, particularly those due to the humoral and cellular immune responses to the viral structural proteins that form the vector particle [18,19]and any de novo virus gene expression [2,20]. Plasmids produced in E. coli contain elements needed for propagation in prokaryotes that are unnecessary, and even deleterious, for transgene expression in mammalian cells [21]. In addition to the required cis-elements, including a prokaryotic origin of DNA replication (e.g., colE1 ori) and a selectable marker, typically an antibiotic resistance gene (e.g., b-lactamase), bacterial plasmids bear uniquely prokaryotic modifications to DNA, such as N6-methyladenine and N5methylcytosine [22]. These may elicit undesirable outcomes in an in vivo application [23]. Relative to mammalian genomic DNA, the occurrence of CpG dinucleotides in prokaryote-derived pDNA is overrepresented [24] and reportedly binds a member of the Toll-like family of receptors, eliciting a T cell-mediated immune response [25,26,27]. In addition, reducing the amounts of detrimental impurities, including bacterial genomic DNA and endotoxins, is critical for production of plasmids used in vivo.
''Mini-circle'' DNA, a recent improvement to the production of pDNA free of prokaryotic cis-acting elements, uses an inducible recombinase to specifically delete the prokaryotic moiety from the bacterial mini-circle shuttle plasmid [28,29]. Mini-circle DNAs have been used in vivo for long-term gene expression in mice, and for generating induced pluripotent stem cells (iPSC) in vitro [30]. Another alternative to pDNA is mini-linear, covalently closed (mini-lcc) DNA that possesses all the benefits of ''mini-circle'' DNA vectors and effectively eliminates the potential for undesirable vector integration events [31]. However, both mini-circle and mini-lcc DNA are produced in prokaryotes. We have described an alternative method for the high-yield production of eukaryotederived, prokaryotic-sequence-free DNA for gene transfer. CE-LiD-vector DNA consists solely of a therapeutic or reporter gene transcription unit and the AAV2 ITRs. The cis-acting elements and trans-acting requirements needed for propagating the parental plasmid in E. coli are eliminated from the CELiD ''rescue'' and replication process. The genetic organization of CELiD-vector DNA, resembles rAAV vector DNA, but differs in conformation. The encapsidated recombinant AAV vector genome is predominantly a unique sequence of linear single-stranded DNA, although self-complementary forms are packaged either by designing a vector genome with a defective ITR [32,33,34], or through natural processes whereby half wild-type-length genomes are packaged as self-complementary DNA [35]. Unlike encapsidated AAV vector genomes, CELiD-vector DNA has no packaging constraints imposed by the limiting space within the viral capsid. In theory, the only size limitation resides in the DNA replication efficiency of the host cell. Infection of a stable cell line bearing an integrated copy of a 2.7 kb GFP vector DNA with a Repexpressing baculovirus can yield up to 60 mg of CELiD-vector DNA from 4.7610 9 cells. This quantity of CELiD DNA represents: 1) $4 million copies per cell, 2) approximately 15-times the mass of the cellular genomic DNA, and 3) approximately 0.36% of the cell mass. Thus, CELiD-vector DNA represents a viable eukaryotic alternative to prokaryote-produced pDNA.

CELiD as a biological replicative intermediate
Autonomous, linear DNA replicons with covalently closed ends (i.e., terminal hairpins) have been identified in a wide variety of prokaryotic and eukaryotic organisms. Examples include certain large DNA viruses of eukaryotes (such as poxviruses and African swine fever virus), as well as various temperate phage of gramnegative bacteria (such as N15 and wKO2 of Escherichia coli and Klebsiella oxytoca, respectively). Additionally, linear plasmids and chromosomes bearing covalently closed ends have been described in Agrobacterium tumefaciens and various spirochete species of the genus Borrelia. Linear replicon-encoded trans-acting factors responsible for terminal hairpin formation have been identified and biochemically characterized. In particular, the ResT protein of Borrelia and the TelN protein of phage N15 have been extensively examined (reviewed in [36,37]). We have demonstrated that a transgene sequence flanked by adeno-associated virus inverted terminal repeats (i.e., CELiD DNA) is maintained within S. fugiperda insect cells as a covalently closed, linear DNA molecule that serves as an AAV Rep protein-dependent replicon. Interestingly, the large AAV Rep proteins share biochemical activities with the ResT and TelN proteins, including DNA binding activity, tyrosine-mediated phosphodiesterase activity and the ability to join cleaved DNA strands [38,39,40,41,42]. We postulate that Repmediated DNA cleavage-joining activity is responsible for generation of the covalently closed DNA structure observed for CELiD DNA.
Perhaps the best-known examples of covalently closed linear replicons in eukaryotes are those of the Poxviridae. Poxvirus DNA replication bears similarities to that of the Parvoviridae (including the dependoviruses such as AAV). DNA replication in both virus families initiates by enzymatic nicking of one of the terminal hairpins flanking each end of the linear genome [43,44]. The nicking event provides a 39-hydroxyl group to prime stranddisplacement replication. In addition to providing a self-priming DNA element to initiate DNA synthesis, covalently closed terminal hairpin structures may protect the linear DNA elements from attack by cellular exonucleases, as well as inhibit illegitimate recombination events. The role, if any, that closed-ended replicative forms play in wild-type AAV replication remains to be determined. Figure S1 Restriction endonuclease digestion products resolved using native and alkaline agarose gel electrophoresis. (A) Native (1% agarose, TBE) and (B) denaturing (0.7% agarose, alkaline) gel electrophoresis. For both gels, CELiD-GFP DNA samples in lanes 1-6 were digested with the following restriction endonucleases. Lane 1: AgeI (monomer endonuclease recognition site is at position 1801, fragment sizes 1801/867 bp).  Table of predicted restriction endonuclease digestion products. AgeI, SpeI, XhoI and KpnI have single recognition sites, whereas NotI and XbaI have two recognition sites in the CELiD-GFP monomer sequence. Digestion products resolved using native agarose gel electrophoresis produced DNA fragments predicted by the rAAV-GFP (and corresponding CELiD) map, listed in the table by fragment length (x, y, or z) in basepairs. The denaturing, alkaline agarose gel products were 2x and 2y nucleotides in length for the single-cut enzymes, whereas the double-cut enzymes produced 2x and 2z length products, and an internal product y that appears the same size on both native and denaturing gels.